Tracking AI Gigawatts, Revenue, Expenses and Deep Dive into Productive Agents

The rapid escalation of artificial intelligence (AI) usage, particularly through productive agents like Grok bot, is driving a significant increase in energy requirements and the construction of AI data centers. This trend is underscored by the continuous growth in AI revenues, highlighting the necessity for a detailed analysis of the current dynamics in the sector to inform investment and business decisions.

Current State of AI Compute Power

According to SemiAnalysis, both Anthropic and OpenAI are projected to end 2025 with nearly 2 gigawatts (GW) of AI compute power, with expectations to reach approximately 5 GW by the end of 2026. Epoch AI's data center tracking provides specific insights, estimating that Anthropic will have around 1.45 GW and OpenAI about 421 megawatts (MW) of compute power by the end of 2026. However, when considering all announced projects for 2026, there appears to be a shortfall of 4.4 GW across these two companies. This gap is likely due to increasing AI cloud purchases from major players like Microsoft, Amazon, and Google, which are intensifying the demand for AI infrastructure.

In a related development, Elon Musk has indicated that SpaceX is ramping up its power and AI data center capabilities faster than its competitors in 2026 and 2027. Musk's comments highlight a growing power shortage in the industry, with Google and Anthropic reportedly leasing AI compute resources from SpaceX. He emphasized that SpaceX has been more effective in bringing AI compute online compared to other companies, suggesting that their infrastructure may be crucial in addressing the current demands of the AI sector.

Financial Dynamics of AI

The financial landscape surrounding AI compute is complex. Anthropic's second-quarter expense estimates suggest that they will operate around 2.7 GW of compute power in 2026, which is notable given that they incur higher costs due to SpaceX's rental prices for data center usage. The realized revenue in the industry is approximately $7 billion per GW per year, which is nearly equivalent to the cost of owning and operating a gigawatt of compute power, estimated at $7.9 billion. This figure starkly contrasts with the projected revenue of $100 billion per GW annually for frontier AI capabilities from companies like Anthropic and OpenAI.

The implications of these financial metrics are significant for stakeholders in the AI industry. For instance, Codex, an AI tool, has reached 20 million users, although it has not yet achieved 20 million developers. This discrepancy indicates a broader trend where user engagement may not directly translate into developer activity or revenue generation.

AI spending patterns also reveal disparities among different user segments. Data from Ramp indicates that the top 1% of AI spenders currently average $7,400 per month, translating to an annual expenditure of $88,800. Forecasts suggest that while token costs are expected to decrease over the next three years, the demand for AI services will continue to rise, further complicating the financial landscape for creators and technologists.

Why it matters

The increasing demand for AI compute power and the corresponding energy requirements have significant implications for creators and technologists. As companies like SpaceX position themselves as key players in the AI infrastructure space, the competitive landscape is likely to shift, influencing pricing and availability of resources. This could affect how smaller companies and individual creators access AI tools and technologies, potentially widening the gap between larger organizations and emerging startups.

Moreover, understanding the financial dynamics of AI compute is crucial for making informed investment and business decisions. As the costs of operating AI infrastructure approach the revenues generated, companies must navigate a delicate balance to ensure sustainability. The trends in AI spending, particularly among the top spenders, suggest that while there is a robust market for AI services, it may become increasingly challenging for new entrants to compete without significant capital investment.

In summary, the current trajectory of AI usage, infrastructure demands, and financial realities presents both opportunities and challenges for creators and technologists. As the landscape evolves, stakeholders must remain agile and informed to capitalize on the potential of AI while mitigating risks associated with its rapid growth.